Intel To Launch Three Nehalem-based CPUs by End of Q408?

[Rob Williams]

From our front-page news:

It's still a bit early to instill confidence in rumors, but this one seems reasonable. Industry-tracking site DigiTimes is reporting that by the end of the year, Intel will launch three different processors based on the Nehalem microarchitecture. The fastest model will debut at 3.2GHz, similar to today's LGA775-based QX9770, while the others will be clocked at 2.93GHz and 2.66GHz.

If history is any indicator, Intel will launch the 3.2GHz part first and label it as part of the 'Extreme' series. After some time has passed, they'll follow-up with the lower-clocked offerings, which are still likely to be priced higher than current Yorkfield mid-range offerings.

As we are already aware, the X58 chipset will launch at the same time as the first processor's launch, which will end what we know as the FSB, thanks to the introduction of the QuickPath Interconnect architecture. As for further specifics, such as pricing and model names, that won't be known until later this year, though leaks are sure to happen before then.

In other news, Intel has updated its processor schedule in will begin to phase out its Core 2 Extreme QX6850 and 6800 after July this year, the sources revealed. Additionally, the performance-level quad-core Q9550 and Q9650 will both phase out in the first quarter of 2009. Intel will leave its quad-core Yorkfield processors for the mainstream market and their life cycle will not end until the end of fourth quarter 2009.

Source: DigiTimes
​

 

[Kougar]

Hm, most of this has been mentioned, but nice to have some sites like DT put their name behind it. I'd link to EXPReview's article on this from a month ago but the link no longer works... or their site for that matter. Might be something on my end though.

The roadmap/chart indicated that the 3.2, 2.9, 2.66 parts were targeted at the Extreme, Performance, and Mainstream segments respectively.

This is actually really great news, because the "mainstream" 2.66GHz part will cost either around $320 or the $512 mark. I am betting it will be ~$320-380... Other chips listed beside the 2.66GHz Bloomfield are the E8600, Q9550, Q9400 to give an idea on the pricing.

The roadmap also appears to show that Q2'09 Intel will be launching higher clocked parts to replace all three of these...

 

[Rob Williams]

I have no idea how I missed that. I knew that the 3.2GHz model was known about, but I don't remember hearing about the other ones.

I guess it's just the excitement building up ;-) I just hope they manage to blow us off our feet once we get them. It will be hard to top the Core launch, that's for sure.

 

[Kougar]

Hehe, things get missed all the time, I'm just a bit crazy and keeping a very close eye on all things Nehalem. I'd be the first to say I'm not always right, but I sure got it right with Conroe... looking to do so again.

From all the leaked info I'm seeing, and especially Anandtech's illicit coverage, I don't think we have anything to worry about!!

They were using early silicon and it still wasn't nearly as hot as I was expecting... I'm very curious if the ease of overclocking the CPU multi verses the QPI bus will make the Extreme chip better worth it's price... I'm not that confident in how far the QPI bus can be pushed, since it works at a leisurely 133MHz.

 

[Rob Williams]

Yeah... I tend to miss a lot of things. For some reason, there is a lot of info my brain just doesn't seem to retain. Sometimes I'll see something and think, "Oh wow, I didn't know that.", then find a post on THESE forums from two months ago where I talked about it.

Regardless, I agree... Nehalem is looking good. As for overclocking, I didn't have a hands-on experience, but I did see what the chip was capable of behind closed doors, and all I can say is that I was really impressed. Things should only improve from this point forward.

 

[Kougar]

Yeah... I tend to miss a lot of things. For some reason, there is a lot of info my brain just doesn't seem to retain. Sometimes I'll see something and think, "Oh wow, I didn't know that.", then find a post on THESE forums from two months ago where I talked about it.

Hah, I am exactly the same way, even if it found the original topic to be very interesting. Is partly why I have so much trouble recalling exactly /where/ I read some specific news report or article. I think I remember where I found the original link though, so will see about finding that ATI GPU aftermarket cooler one.

Regarding the overclocking, I know the chip itself has decent headroom, at least. But there might be a huge discrepancy between overclocking via QPI frequency verses overclocking via CPU multiplier... apparently that makes for a significant difference with AMD Phenoms if you have a specific southbridge chipset right now for example.

 

[Rob Williams]

I have no idea about the QPI... but it should be interesting. The overclock I saw was using a default QPI, but high multi.

If QPI is actually difficult to OC... then that's going to make the Extreme editions much more tempting, or required...

 

[b1lk1]

1. I do not understand the hype about Nehalem. I understand it is a new architecture, but it isn't any significant speed or power leap.

2. What about the fact that only extreme edition CPU's are going to be allowed to overclock? Stories are everywhere about the reports of the mainstream CPU's being locked out of overclocking.

 

[Kougar]

1. I do not understand the hype about Nehalem. I understand it is a new architecture, but it isn't any significant speed or power leap.

It is as significant a leap as Conroe was in terms of performance, while maintaining the same power envelope we presently have. If that's not enough performance to get excited about, then I don't know what is.

2. What about the fact that only extreme edition CPU's are going to be allowed to overclock? Stories are everywhere about the reports of the mainstream CPU's being locked out of overclocking.

I've said this elsewhere, but perhaps not on Techgage... I firmly believe it's just a textbook case of FUD.

Not to mention... Anandtech had no issues playing with the QPI frequency, the motherboards just weren't functional yet so they couldn't raise it.

 

[b1lk1]

Unless it is a full retail sample (which it OBVIOUSLY isn't) then you cannot use that as an example. There have been stories released by Intel stating this. It isn't fud....

 

[Kougar]

Right up until official Intel documents were leaked, it was believed and reported that there would only be a single Bloomfield chip, notably an $1k+ Extreme. Now we know otherwise... I have seen nothing official from Intel claiming there will be no overclocking, in fact some Intel employee has refuted those claims. Days after TGDaily posted news about the no overclocking, they changed their minds: Link

I was convinced the no overclocking rumors were FUD before Anandtech got their mits on a Nehalem chip. They didn't overclock via multiplier, they tested using QPI instead. So my point is, regardless of Anandtech's preview I believe "no overclocking" to be FUD. Want to place bets?

Lets say I am wrong and overclocking isn't allowed. How long do you think it would take people to discover pin mods? There will be 1366 pins to play with, I am sure several of those will alow higher QPI links. It would be easy for a motherboard maker to "software" pin mod a CPU.

 

[Rob Williams]

Not to mention... Anandtech had no issues playing with the QPI frequency, the motherboards just weren't functional yet so they couldn't raise it.

The problem is how much the QPI frequency will have to be increased to see a "good" overclock without touching the multiplier. It's going to be a sticky situation if Intel does lock out overclocking, but no one knows that right now except them.

It'd be foolish. As I've said before, if people had no option to overclock except by getting an expensive CPU, then people will simply not overclock anymore. There are few people who would pay upwards of $1,000 for the sake of enabling something not-so important, especially when the stock speeds of the chips should be kick-ass as is.

 

[b1lk1]

The same crop of "look at my scores" will flock to the $1K+ chips like a moth to a candle just to show how impressive they are. Otherwise, there is almost no need for the rest of us to overclock since there is little gained in real world usable performance. I would not be surprised in the least to see them lock it out.

Kougar, my point is that we can all have out opinions and noone is right yet, including you. I'd place a bet on either right now because it is 50/50 and since AMD is of no real competition, this is an ideal time for Intel to do this and not lose a single bit of their customer base. Why continue selling cheap CPU's that clock up to their expensive brothers when you can just force people due to lack of competition to pony up for that extra performance.

PS: Anandtech is not Intel as well. I take any and all comments about this with a larger than life grain of salt until we see real world retail samples, not these cherries that Intel is famous for handing out before launches. If you don't believe Intel does that, you are kidding yourself.

 

[Kougar]

Oh, I've been wrong plenty of times, quite a few actually since I tend to speculate above and beyond the facts. But I've also read every scrap of info I've been able to dig up. With Conroe I was pretty spot on, although there was a general belief was Intel was "cooking the books" and cherry picking at the time.

By dividing the platform into a low cost LGA1160 verses a performance LGA1366 sockets Intel already is going to ensure people won't be taking cheap CPUs and overclocking them beat the performance of their flagship chip. There are significant platform differences that I do expect to see show in performance tests. If there are only three CPUs, the cheapest being around the $334 ballpark then Intel has already won here, and has no reason to push things further by limiting overclocking in my opinion. $334 will be the cheapest chip until Q2'09

I think it is fun to try and figure all this out, so don't get the wrong impression and think I'm trying to be a know it all or something. However I will say I do remember the cherry rumors were strong before Core 2 Duo launched as well... if anything, the launch day product performed even better and I expect history to repeat again here once they get all three memory channels working, BIOS's tuned, and non-reference boards made.

 

[b1lk1]

Believe me, I am also not trying to be aggressive. My main point is that Intel is a business run by people that want to make money. They had to forcefully remove AMD from the high end fully at any and all means possible. Now that that job is complete, the time to start jacking up profits is coming. They certainly do not care about their customers on a level of genuinely wanting us to have a cheap CPU because it is the nice thing to do. It was a business decision and a very smart one at that. The time is ripe for them to put the screws to the consumer and they have an easy excuse to do it in stating there is some reason we should not be allowed to or need to overclock anymore. I strongly feel that the easy overclocking days are coming to an end. Sure there will be pin mods and all other kinds of mods that were once the only way to do it, but it makes ALOT of business sense to stop selling cheap overclockable chips.

 

[Kougar]

On the point that Intel is a business whose primary goal is to make a profit, I quite agree with you. I don't feel the time is ripe for them, as you put it, because should they try anything right now then it gives AMD's legal case all the grounding they need to prove Intel is still abusing it's majority market share.

You may be right that the days of easy overclocking are at an end, but I don't think overclocking a Nehalem chip will be any harder than an AMD chip we already have.

Interestingly enough, here's some fresh news. Looks like the 2.66GHz Bloomfield will retail for around $284. Link

 

[Rob Williams]

I talked to Intel briefly about this earlier today, and while they are still not saying much, we'll find out a lot more at next month's Intel Developer Forum.

I am a bit more confident now that no one will lose the ability to really overclock, it's just a matter of how far the chips will be able to be pushed.

 

[Kougar]

I am a bit more confident now that no one will lose the ability to really overclock, it's just a matter of how far the chips will be able to be pushed.

That is exactly what I am thinking... the chips will overclock great, but the QPI bus won't be able to overclock so well. If so then having the ability to change the CPU multi should really differentiate their Extreme chip... will have to wait and see I guess.

 

[Rob Williams]

That is exactly what I am thinking... the chips will overclock great, but the QPI bus won't be able to overclock so well. If so then having the ability to change the CPU multi should really differentiate their Extreme chip... will have to wait and see I guess.

Aye... right now we can't do a single thing but wait. Though I was told we will learn a lot more at IDF, I'm not that confident that non-Extreme chips will be there. After all, those are always the first to launch. Just a waiting game here-on out...

... unless of course some Chinese site happens to stumble on an extremely early sample, like they always seem to do.

*taps foot*

 

[Kougar]

Plenty of samples have been floating around... already been some random threads on the XS forums with random benches to complement the array of news reports. I've honestly not followed the XS threads closely though, they keep running benches for whichever OS I'm not presently running. It gives new perspective when you can compare your own best 24/7 OC to the results

What is being reported (and I believe to be accurate, not disguised rumor) is that Intel will be launching their Bloomfield chips this fall... so that should mean all three of them. LGA1160 won't even be seen until Q2'09. Some sites report even a bit later, but I'm not falling for that one.

What is interesting is the QPI bus looks to be "Quad Pumped", just like the FSB. I'm not sure if this is just because software aka CPUz hasn't adapted yet, but they always show the 133Mhz with an effective 533MHz speed. This was identical to P4 Northwoods... I get the funny feeling come Q2'09 we will be seeling 200MHz QPI links for a total of 800MHz, just like later editions of the Northwood P4s...